An Efficient Hierarchy-Based of K-Means Clustering Algorithm

An efficient hierarchy-based of K-means clustering arithmetic is presented. Grounded on statistical mechanics, the partition matrix element (membership probability) on the basis of the traditional of K-means clustering is changed formally and a Lagrange multiplier controlling the clusters number is introduced. In this way, for a given dataset, the result will get different clusters number when the Lagrange multiplier is not the same. The method is tested on a synthetic data set. The result demonstrates hierarchy feature and more satisfied with the accuracy of the cluster, and more efficient to initialize the cluster centers.